File size: 1,786 Bytes
d408ee2 90a810f d408ee2 dbac39f d408ee2 367ff93 d408ee2 f9aad7c 5260ab4 d408ee2 f9aad7c d408ee2 dbac39f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 |
---
language: en
tags:
- sentence correction
- text2text-generation
license: cc-by-nc-sa-4.0
datasets:
- jfleg
---
# Model
This model utilises T5-base sentence correction pre-trained model. It was fine tuned using a modified version of the [JFLEG](https://arxiv.org/abs/1702.04066) dataset and [Happy Transformer framework](https://github.com/EricFillion/happy-transformer). This model was pre-trained for educational purposes only for correction on local Caribbean dialect. For more on Caribbean dialect checkout the library [Caribe](https://pypi.org/project/Caribe/).
.
___
# Re-training/Fine Tuning
The results of fine-tuning resulted in a final accuracy of 90%
# Usage
```python
from happytransformer import HappyTextToText, TTSettings
pre_trained_model="T5"
model = HappyTextToText(pre_trained_model, "KES/T5-KES")
arguments = TTSettings(num_beams=4, min_length=1)
sentence = "Wat iz your nam"
correction = model.generate_text("grammar: "+sentence, args=arguments)
if(correction.text.find(" .")):
correction.text=correction.text.replace(" .", ".")
print(correction.text) # Correction: "What is your name?".
```
_
# Usage with Transformers
```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("KES/T5-KES")
model = AutoModelForSeq2SeqLM.from_pretrained("KES/T5-KES")
text = "I am lived with my parenmts "
inputs = tokenizer("grammar:"+text, truncation=True, return_tensors='pt')
output = model.generate(inputs['input_ids'], num_beams=4, max_length=512, early_stopping=True)
correction=tokenizer.batch_decode(output, skip_special_tokens=True)
print("".join(correction)) #Correction: I am living with my parents.
```
|